Bob Smith Termination: Is it Correct for Ethernet?
Table of Contents
From the time my company received its first job involving Ethernet, we always “Bob Smith” with gusto. It wasn’t until I was asked to write an article on Ethernet grounding for Signal Integrity Journal that I ever gave Bob Smith termination a second thought. As I started to dig on the internet, I started to find some objections to Bob Smith termination. Some of these objections were purely conceptual, while others were backed up with data.
I was surprised to see that this was so controversial; many designers claim Mr. Smith is flat-out wrong, while others always follow Mr. Smith and never seem to have problems. So who is correct? Is this one of those cases where application notes push bad design advice and everyone follows along, or is this legitimate design guidance that is taken out of context?
Unfortunately, application notes do an outstandingly horrible job of explaining Bob Smith termination, if they try to explain it at all. I’ve never seen an application note that doesn’t advocate Bob Smith. But as most senior designers have stated on this blog, application notes are not to be trusted with impunity. Let’s look at this issue around Bob Smith termination in more detail.
Originally patented by Bob Smith, this refers to a certain set of resistors that provides termination for the center tap of a common-mode choke that is used in Ethernet routing. When routing between an Ethernet PHY and discrete magnetics circuitry, this termination scheme is used to ground the center taps of the transformers used in the magnetics circuits. Note that this is used in Ethernet routing to provide a sink for common-mode noise that may transfer from the PHY side to the connector side.
Bob Smith termination uses four 75 Ohm resistors (2 for Rx, 2 for Tx) and a capacitor to provide an impedance-matched path back to a ground point in the system. Depending on which application note you read, you’ll find that the ground point ranges from chassis ground to analog ground, although this is a different signal integrity issue that relates to grounding and planning mixed-signal return paths.
The image below shows a magnetics circuit with Bob Smith termination on the transformer center taps for a 100 Mbps Ethernet link. The Bob Smith termination scheme is outlined in red. C3 ranges anywhere from 1 nF to 4.7 nF, depending on the system’s bandwidth.
If we look at the above circuit, the need to terminate the center tap to ground seems to make sense for common-mode noise reduction. By diverting some common-mode emissions to ground, you’ve effectively increased the system’s overall common-mode rejection ratio (CMRR). This means the return loss back to ground should be as low as possible. This is where the objections to Bob Smith termination come into play.
I’ve never met Jim Satterwhite, nor had I ever heard his name before looking into the achievements of Bob Smith. Jim is probably the most cited author to claim that the Bob Smith termination scheme is not optimal, and that a different termination scheme should be used. You can read his article on the topic here. His solution is simple: use 52.3 Ohm resistors instead of 75 Ohm resistors.
Very simply, Satterwhite’s objection is that Bob Smith’s termination scheme would only be optimal if the differential impedance of a typical UTP cable was 145 Ohms. Obviously this is much higher than the differential impedance of 100 Ohms used in Ethernet cables. Satterwhite measured the impedance of a single differential pair in a UTP cable versus other configurations and retrieved the specified characteristic and differential impedances, although I don’t think he realized what he was measuring.
Satterwhite then compared return loss values for his scheme and the original Bob Smith termination scheme, and he found that his proposed scheme provides ~10 dB less return loss for common-mode currents entering the system ground. This is clearly an improvement; less common-mode noise is reflected and we’d expect less common-mode EMI from this section of the system and from a UTP cable itself. What happens next depends on keeping the return path for this noise away from ferrites above the system ground, which has raised its own set of objections from the SI community.
If you read through online PCB design forums, you’ll see other designers quoting a study by Royce Bohnert. He showed that the Bob Smith termination scheme (75 Ohm resistors), Jim Satterwhite’s modified scheme (52.3 Ohm resistors), and no termination at all did not appear to produce any difference in return loss measurements. The original link to Bohnert’s presentation has become insecure, but you can download a PDF of the original presentation from here.
Despite the objections, I don’t see a reason to not use a termination network with this topology, regardless of whether it uses 52.3 Ohms or 75 Ohms. When in doubt, it doesn’t hurt to run a simple simulation with your choke and proposed termination method. You may find that Satterwhite’s assertions are correct and you’re better off using 52.3 Ohms instead of 75 Ohms in your termination network. With the right design tools, you can run SPICE simulations with specific component models directly from your schematic and determine which termination network is right for you. Don’t forget to include the capacitor in your simulations!
Whether you plan on using Bob Smith Termination or another method, the schematic design and PCB layout tools in Altium Designer® give you everything you need to create advanced electronics. You’ll have a complete set of features that makes routing high speed impedance controlled designs an easy task.
Altium Designer on Altium 365® delivers an unprecedented amount of integration to the electronics industry until now relegated to the world of software development, allowing designers to work from home and reach unprecedented levels of efficiency.
We have only scratched the surface of what is possible to do with Altium Designer on Altium 365. You can check the product page for a more in-depth feature description or one of the On-Demand Webinars.